Supply Chain Data Quality and Resilience Explained

What if the biggest risk in your supply chain isn’t geopolitical shocks or new regulations, but the data you trust every single day?

This week, I’m joined by Andy Kohm, co-founder and CEO of SCIP, a supply chain intelligence platform built to clean, connect, and operationalise data across ERPs, PLMs, control towers, and the spreadsheets nobody admits to using. Andy has spent more than a decade wrestling with the messy reality of supply chain data, and his insights couldn’t be more relevant as volatility rises and digital transformation hits its limits.

In this conversation, you’ll hear how bad data quietly drives bad decisions - from inflated lead times to unnecessary expedites to risk scores that collapse under scrutiny. We break down why most organisations can’t agree on something as simple as the “source of truth,” and how that single failure cascades into higher emissions, higher costs, and planners who simply stop believing the system.

You might be surprised to learn how often companies pay 10x for components they could have sourced at the normal price - simply because the underlying data was wrong. And we dig into where AI can genuinely help today (contract intelligence, grunt-work automation) and where it’s still pure theatre without clean inputs.

🎙️ Listen now to hear how Andy and SCIP are reshaping the future of resilient, sustainable, data-driven supply chains.

Podcast supporters
I'd like to sincerely thank this podcast's generous Subscribers:

• Christian Sederberg
• Kristal Maharaj
• Alicia Farag
• Kieran Ognev

Transcript

Andy Kohm: 0:03

It's all about, ensuring that we make decisions on the right information. It's not sexy. It's not, you know, glorious. But you know, if your data's not right, you throw AI on it, it makes bad decisions faster.

Tom Raftery: 0:18

Good morning, good afternoon, or good evening, wherever you are in the world. Welcome to episode 96 of the Resilient Supply Chain podcast. I'm your host, Tom Raftery. Some supply chain risks announce themselves with drama, a factory fire, a port strike, a geopolitical shock, but the most destructive risk almost never does. It sits quietly in your systems, buried in mismatched part numbers, contradictory lead times, orphaned BOMs and spreadsheets that haven't been touched since the pandemic. And while companies talk about AI, control towers, and digital transformation, the truth is painfully simple. If your underlying data is wrong, every smart system you built on top of it simply helps you make the wrong decision faster. Meanwhile, volatility hasn't eased. It's multiplied. Lead times are drifting again. Tariffs shift overnight. Forced labor rules are tightening and scope three pressures now demand part level visibility most organisations still can't provide. The gap between what leaders think their supply chain looks like and what it actually is has never been wider. So How do you run a resilient, low risk, low carbon supply chain when your core data infrastructure is fragmented, outdated, or flat out wrong? To unpack this, I'm joined by someone who's been tackling this problem for more than a decade. Andy Kohm, co-founder and CEO of SCIP. A platform built to clean, connect and operationalise supply chain data so teams can make decisions based on truth, not tribal knowledge. From master data to AI, from sourcing risk to emissions, this conversation goes deep into what leaders need to fix first, and why everything else depends on it. Andy, welcome to the podcast. Would you like to introduce yourself?

Andy Kohm: 2:21

Yeah. Thank you, Tom. My name's Andy Kohm. I'm the co-founder and CEO of SCIP.

Tom Raftery: 2:27

Okay. And for people watching or listening, Andy, what is SCIP?

Andy Kohm: 2:34

Yeah, so SCIP is a supply chain intelligence platform. You know how, our supply chains are generating more and more data every day, but it gets scattered across all of our different systems from like an ERP system to a SNOP system, even in Excel spreadsheets and emails. What SCIP does is it brings all this data together into a single platform where people can then view and use this data. It takes that messy information, it cleans it, it validates it, and it enriches it and allows everyone in the organisation to have access to that information.

Tom Raftery: 3:11

And Andy, what's your path into SCIP? Because co-founder, what made you decide one morning, gosh, I gotta set up SCIP.

Andy Kohm: 3:22

Yeah, so SCIP was actually more than a decade in the making. I didn't start my career in supply chain. I actually started as an R&D engineer in medical devices, Yeah, got really frustrated with not being able to find suppliers, and trying to figure out where does this data live? Then as my career progressed, I started working more and more in the supply chain, part of the company and realised that I'm not the only person with that problem. Many people in, the companies having problems with, knowing what data do we have? Where does it live, can I, believe it or not, and how do I use it? So that's where SCIP came from, was, you know, one day just being so frustrated that I was like, I need to solve this problem.

Tom Raftery: 4:03

Okay. If you look at, OEM. Versus contract manufacturer supply chains today. What's the bit that's most quietly broken but doing the most damage, do you think?

Andy Kohm: 4:18

I think it all starts with the base data, the underlying data that everyone is making decisions on. Everyone assumes it's correct. If it's in the system, it must be correct. It must be being maintained, so I should be able to use that for my calculations or any sort of decision making I want. And what we found is, with these companies with so much data, and so vast data, many of the data is contradicting each other in different systems, is not being maintained or updated, was put in once real quickly. And so that data everyone's making decisions on is flawed. So which means decisions are flawed, which means the outcomes are flawed. So we're starting with one of the most important parts is trying to fix that underlying data that everyone's making decisions on.

Tom Raftery: 5:06

And what's one recent disruption that exposed a blind spot everyone thought they'd already solved?

Andy Kohm: 5:16

I think one of the items that came up was, you know, during COVID there was a lot of these, disruptions to the supply chain. So everyone started focusing on, is the product available? What are my lead times? How do I get it? They did that once. And then, once COVID went away, they stopped maintaining that data, which is very important. So now they're finding out that they go to order from a supplier. The lead times have, expanded and now they're running behind. So they have to expedite and rush shipment, or they have to pause their manufacturing lines.

Tom Raftery: 5:54

Okay. If you could ban one KPI in supply chains tomorrow, which one goes and why?

Andy Kohm: 6:04

Wow, that's a difficult question. I think I always start with a KPI if the value doesn't cause you to create a reaction then it's not really a KPI. You know, so a lot of people are tracking this data. It goes south. They say, oh, well let's just watch it. Right? So I think a lot of these vanity KPIs on, how many suppliers do we have? Or, I would say cost reductions, because a lot of times cost reductions come at a cost of, resilience, risk. So I think a lot of people are really focused on cost reductions when it should be focused on a balanced supply chain between risk, cost, and complexity.

Tom Raftery: 6:50

And we often hear about clean, connected data. What does that actually look like at the part or BOM level?

Andy Kohm: 7:01

Yeah, so I think, at part level it really shows you a universal look at the part. So this goes from attributes, you know, the key items that someone may be in R&D or sourcing needs, the data sheets, all the data you need. But it also goes through the BOM level. What parts is this used in? What finished goods, what assemblies goes into the warehousing? How much do we have in the warehouse? How much do we have on purchase order? You know, and so I think a universal view is actually looking at a part from all the various systems and all the different functions and brings it together so someone can see it in one place and it aligned to make sense.

Tom Raftery: 7:47

Okay. And where do most organisations master data processes fall over? Would it be creation, change, use somewhere else?

Andy Kohm: 7:57

Unfortunately most of the software's really created today to manage the data. So was very good at collecting data, putting it into the system, for a single function and keeping it there. So I think where it really falls down is the ability to connect those systems together and maintain the data. So where the data breaks down is those systems a lot of time allow you to put in, strings of values without any guide rails. So we've seen when people are putting in attributes just to take something simple like a voltage, they'll put, voltage and they put the unit measure in the title. The next person comes in, puts voltage and puts the unit of measure with the value. The next person says voltage in the value. And so with these systems, it treats that as three different entries into the system that are unrelated. So now when you try to use that data, the same record of a 10 volt. You know, part will not come up together. They'll come up in three different section buckets and you can't really see the universal view of what you have.

Tom Raftery: 9:06

Okay. And what would you say is the biggest myth then about control towers that you run into with execs?

Andy Kohm: 9:15

Once again, it goes back to bad data in bad data out. So when you're looking at a control tower and if your data isn't consolidated or aligned, they'll get a improper view of the system. The same holistic view, but if the vendor's in there multiple times, you're not gonna see the total spend by a single vendor 'cause it got broken up. If the part's in there multiple times, you're not gonna see the demand total for that part. So I think when you're talking about control towers, it ties back to having good data feeding into them and then being able to show what matters, the right KPIs

Tom Raftery: 9:53

Let's go hypothetical here. If you inherited a Brownfield tech stack, what's the first 90 day, no regrets move to reduce chaos?

Andy Kohm: 10:06

First, I would map what is the source of truth for every data system. So the problem we have today is they can have a price in an ERP system and a price in a PLM system that is totally different for a part, and no one knows which of those sources is actually the master data to the belief. So the first thing I would do is come in and take all the systems and make sure that there's only a single master data point for all of the data that they're using, so that people know what to use as the reference point, and then roll that out to the all the other systems so people are doing manual entry across all the various systems for the same data.

Tom Raftery: 10:50

And then how would you reconcile price versus lead time versus risk in a way a planner will believe and act on?

Andy Kohm: 10:59

Yeah, I mean, that's always the balance, right? If I'm looking for the lowest price, it might have the longest lead time and it might be from the riskiest region, to purchase it from. I think when I'm thinking about that risk is different for every company. And it's also probably different for every type of part you have. So you probably need to come up with some sort of metric score to figure out where's the balance between those. And also, if you can't really trust your lead time, that shouldn't be the key metric to go off of. So which part of the data is the most trustworthy for you to reference?

Tom Raftery: 11:41

Alright, now, obviously AI has been in the news a lot the last couple of years since the launch of ChatGPT particularly, where does it genuinely help today and where is it still theater?

Andy Kohm: 11:55

Yeah, I think this is, the trillion dollar question that's out there in the world today about what can AI do? Where does it go? And, I think your question really points to a article I just read that 40% of AI initiatives will fail in the next year and a half. And now the question is why.

Tom Raftery: 12:14

Mm-hmm.

Andy Kohm: 12:14

You know, and I, I think it goes back to what your question was. When should I use AI and when should I not use AI, especially generative AI, which is what's in the news today. I think it's really good right now at reading contracts, pulling out the key terms, so like looking at MSAs, looking at your key terms, pulling them out so that you have 'em and you can reference 'em versus someone reading through the contract to pull them out. I think it's really good at fat answers about general data. Where I think it's failing today and we keep on seeing it and I see it as well, is the deep analysis and anytime, you know, a lot of numbers and numerics come in. You ask it to do a calculation, you ask it to give you, a, a real big data, which is the best out of these items, you know, it fails. Also, I always tell people, don't use it for any, process which can tank your company. You know, I view it if, if you're okay with being 90% correct, you can use AI if you need to be 99.9% correct. It's probably not the time to be using AI today.

Tom Raftery: 13:29

Okay, but that's generative AI in particular i'm guessing because other forms of AI tend to be better.

Andy Kohm: 13:37

Yeah, I mean, AI's been around since really the 1950s. Most people just didn't realise that it was in different forms of AI. The most common is machine learning. Machine learning is very good, very robust, and is very good at calculations. The new hype that everyone's seeing is really generative AI, which is, a huge leap forward to what AI can do. But also right now, still in it's infancy, stage with capabilities and people's understanding on how to use it correctly.

Tom Raftery: 14:09

What would you say is a good human in the loop design for a sourcing decision? So that it's faster and safer?

Andy Kohm: 14:19

With AI, let AI do all of the groundwork, let AI go, do all the grunt work, right? But I think at the end of the day right now, there should be a human in the loop that is able to verify and validate the information. So a lot of times when I use, AI to go get something, one, I ask it to check itself. A lot of times when you run something, it'll give you one answer, and my next question is validate that the answer you gave me is correct. And a lot of times it'll fix itself. It's, it's amazing how well that works. but when I use AI as well, I always ask it to give me the underlying data that it used to make the decision. And therefore, when I see it, I can look at the results or the conclusion it came to, and quickly at least see, is it factually correct. Does it make sense? So that's how I use it with a human interaction. And I think today it should be, AI assisted, we call it AI powered in our system as well, where a human is still in the loop, just ensuring that the data and the calculations and the information is correct.

Tom Raftery: 15:26

Okay, good. And with the likes of forced labor and due diligence rules tightening now, does operationalising compliance look like on a Tuesday afternoon?

Andy Kohm: 15:42

Yeah, I mean, all of the compliance requirements are becoming more and more strict, more and more data intensive. I still don't know how, and without going and visiting your suppliers and all their sites to understand whether or not they use, forced labor, and unfortunately we keep on seeing it in the news. The biggest, companies in the world keep on getting hit with these manufacturers using some sort of forced labor, child labor. Unfortunately the best thing you can do today is, have your companies sign a compliance, document that says they do not use it. How much weight does that really carry? I don't know, but at least you asked the question. You can have, AI once again, go and search for any news articles or, lawsuits against the company with it as a background tech, and that's probably a pretty good use of AI. It would probably be able to find that pretty quickly without you searching. So that might be a secondary measure just to, look at like red flags.

Tom Raftery: 16:43

And how do you trace risk at the component level when supplier disclosure is patchy at best?

Andy Kohm: 16:53

Yeah, I think risk at a component level. A lot of people think of it from a single viewpoint, When I think of risk at a component level, I think of it the whole way through your supply chain, your production plan, even your sales. So a 5 cents capacitor that goes into 30% of all of your revenue parts is a high risk component, no matter where you're buying it from, how many you have. So you want to be, a lot safer, with that component. Carry more in inventory, have more sources, versus, a very expensive component that is used in 1% of your, parts is probably not as risky, no matter where you buy it from, from a risk standpoint, it might be risk to be able to source. But from a company standpoint, it's not a high risk to your company, And so you can take into account, what are the lead times of it? How many sources do you have? Is all of your sources for that component coming from a single country? I mean, these are all things that should be, added to a risk score of a part.

Tom Raftery: 18:09

Okay. And what would your take then be on balancing agility with auditability when tariffs and routes change overnight?

Andy Kohm: 18:21

Yeah, what, you know, really what SCIP is paid for is to allow you to have that agility and because all of the data has been connected and being maintained, it doesn't matter how many components you have, it doesn't matter how many backups you have. And it's all connected, that it can roll up the whole way through a single BOM, through your total, product line through your entire company and give you that data. So if tariffs are applied to Vietnam, SCIP can quickly show you what impact would that have across your entire revenue stream based on your sales plan, based on a single component, based on all of the components. In addition, it can provide opportunities or alternatives so that you could source with the same specifications that are not from Vietnam. So it can also give you those alternatives to help you avoid that tariff.

Tom Raftery: 19:14

Okay. And where then does cleaner data cut emissions fastest expedites avoided mode shifts, scrap waste, or inventory right sizing?

Andy Kohm: 19:28

Yeah, with the right data and the right information, one of the biggest, contributions to emission is these rush shipments where instead of using a very economical, low emissions approach to delivery, you're expediting it by air freight. You're expediting it by a rush shipment, right? And so one, you're doing smaller batches, packaging's probably, a lot more waste. Two, you're not using the most economical, mode of transportation. So with, good lead times, good production plans, it helps cut off that information. The other part with data, if you understand it, correctly, is having alternative parts that come from, better sustainable manufacturers, better locations, closer to your manufacturing plants. All of these can help you cut, emissions and, help you be more sustainable without really affecting your, or potentially even helping your cost and your, manufacturability and your risk scores. You know, it's just getting that data, finding that data takes a lot of time. Especially on those long tail low cost components that no one really wants to spend time on.

Tom Raftery: 20:42

Hmm. Okay. And is there any one simple carbon aware rule that you've seen work? You know, like if lead time is greater than X, never air freight or anything similar?

Andy Kohm: 20:56

Well, it depends on where you are, right? You know, I'm in the US and I know that takes about six weeks for a boat to go from China to the US. Then you add in the ports, then you add in going through customs and you add in the ground transportation. but now you're probably looking at eight to nine weeks if you're buying something from China and doing manufacturing in the US. So, I'm not sure it's really about the lead time. It is more about planning.'cause if it's nine weeks there and the manufacturer has, 10 weeks of lead time, now you're looking at having to order 19 weeks ahead and balancing that. If a stack up and then it, then the next step goes into how well do you, trust your demand plan. You know, the farther out you have to buy this from. If you buy 50 weeks in advance, your sales forecast is probably not really sure on what the sales are. Someone's just waving a finger in the air saying, I think it's this. Now if you can shorten that to, I only have to buy 20 weeks in advance or 15 weeks in advance. have a lot better firm number on how many you really need to buy versus just, purchasing in advance and no one wants to run out of products. But what does everyone do? They over purchase.

Tom Raftery: 22:13

How do you get planners to trust a system after it's been wrong once?

Andy Kohm: 22:20

I always view with any project you need the early wins. The early wins start building the confidence. So you have to figure out where can you start building that trust. Once you have trust, people are willing to understand when something doesn't go right. If you start out with something very complicated and right away it gives someone a bad data point, they quickly dismiss it as, oh, the system is always wrong. So we always start with, what can we do that gets someone coming to the system over and over again and using it and making their job easier before we start rolling out the more advanced systems where they might have an error or every once in a while might give something that's slightly off and people don't discredit it. They just understand that it's still better than what a human would do. So anytime we use these phones, we expect it to be a hundred percent correct. When we get it from a human, we're probably happy if it's 80% correct. And so you have to get people to understand that, you know, systems still aren't a hundred percent perfect, but it's doing a better job than if someone was manually cutting and pasting through Excel spreadsheets, making errors with the copy and paste, putting things into the wrong column, and so it's building a trust first before a error happens where someone quickly discounts the systems.

Tom Raftery: 23:43

And obviously, when you're moving people like that, change management becomes quite a topic and quite a chore. What's one change management move that always pays back and, and one that never does?

Andy Kohm: 23:57

Change management is one of the hardest things for companies to do. Moving forward, going, changing things and going forward. The brutal way of doing it is make sure you don't have a safety net. You know, you have to move forward no matter what. Unfortunately, that can be catastrophic if it doesn't work out. If you don't have it, it's the one-way door versus the two-way door. The problem people have and a lot of times with change management across the board is they don't get the buy-in ahead of time. They try to force it down on everyone without educating or getting the buy-in of why it's necessary and why people should be doing it.

Tom Raftery: 24:37

And tell us about a moment where better data changed a decision.

Andy Kohm: 24:43

Yeah. a lot of what we're doing is looking for. errors, validations, and augmentation of the data. So one of the items that we had with better data was, back at the end of COVID, one of our customers was facing a critical shortage and they couldn't figure out where to buy it or how to buy it or who else has it, or, alternative parts in it. So they were paying 10 x what they should be on a gray market just to make sure their production lines didn't fail. We were able to come in and, you know, show them alternatives, parts that work as well as, paying and give 'em, distributor inventory. So as soon as it came in stock, they were able to buy it at a normal price. When you're talking about better data, I also talk about complete data of being able to see information that you might not have that would be very valuable to help make decisions and also move the company forward.

Tom Raftery: 25:43

For leaders staring at Legacy ERPs and spreadsheets, what should they modernise? What should they layer on, and what should they just leave alone?

Andy Kohm: 25:55

I always say leave alone, anything that's working. I think where you're going today, now is upgrading an ERP, upgrading a PLM, even, you know, I know SAP is trying to force everyone to go to their cloud product, and they're giving them years to make that transition. That should tell you how painful and hard it is to do a transition of those really core systems. You know, once again, they're very good at managing the data. And now a lot of people are trying to change it, so they're also helping you use the data. I think there's a new class and type of software that's coming in saying, no, those systems are correct. People know how to use 'em. They work. All your data's in there. Don't change it. Keep it good, working how it works and use other software that layers on top of all your systems that does the more you know AI, the new calculations, the new technologies that these legacy systems didn't have, One, it saves you a ton of cost from upgrading. Two, it probably will save you a ton of headache where you've changed something core that everyone's used to, and now they have to not only learn a new system, but you threw all these new capabilities into that system that clouds, the functionality or how someone should use it.

Tom Raftery: 27:22

Right. And let's say you were a CTO for a day at a global OEM. What's the first principle you'd write on the wall?

Andy Kohm: 27:31

I'm going back that, the core clean data makes the entire company run, so. So to me it's all about, ensuring that we make decisions on the right information. It's not sexy. It's not, you know, glorious. But you know, if your data's not right, you throw AI on it, it makes bad decisions faster.

Tom Raftery: 27:55

Okay. Right. Okay. And what's the next capability planners will take for granted in two years, three years, four years, that feels futuristic today.

Andy Kohm: 28:08

Yeah. I think right now planners are going from sales forecast and trying to figure out from that sales forecast, what part should I buy? Where do I buy 'em from, and how many do I need? You know, I think a lot of people assume, an ERP system is able to do that, very easily the more complex. But we start digging in and you have five components that you could be using or, the part that you need, going to six different manufacturing locations. Each one uses a different one of those parts or have a different preference. It starts getting very complicated to really understand your rolled up demand. And how you can consolidate that to have lower price, better outcome, lower emissions, you know, 'cause you're fighting against preferences. I think in the future, you know, I hate to say back to the Staples easy button. You know, systems are now gonna be able to go the whole way from the, procurement or even start at the sale and automatically help you figure out what you should be buying, when you should be buying it. And what's the best mix to purchase..

Tom Raftery: 29:20

Okay. Where's all this going? Where do you see the industry going in the next five, 10 years?

Andy Kohm: 29:30

I see it really going towards resilience and lean systems versus focusing on cost. You know, it used to be, just in time, the lowest cost, consolidate everything to the fewest number of suppliers to get buying power. Geopolitical risk, global warming, some environmental risk, always gonna be around for a while. So companies really need to have systems that in, you know, a drop of a dime is able to give 'em an alternate path to continue. So I think a lot of these static systems that are heavily, scrutinised and teams spend a lot of time coming up with the supply chain plan are gonna go by the wayside. And these quick, nimble systems that are able to give you alternatives very quickly is gonna be the path of the future.

Tom Raftery: 30:23

And with that in mind, if you could have any person or character, alive or dead real or fictional as a champion for that, would it be and why?

Andy Kohm: 30:36

Ah. So many, I go back to, the really good engineers of the past who did not take the status quo, like Leonardo da Vinci. No one who just, said, I don't care if you state this is not, true. you tell me it can't be done. You know, I think you really need those people who are going to have an idea and really stick by it and, keep on being the champion of it till people buy it.

Tom Raftery: 31:09

Okay. Fair enough. Cool. We're coming towards the end of the podcast now Andy, is there any question I didn't ask that you wish I did? Or any aspect of this we haven't covered that you think it's important for people to be aware of?

Andy Kohm: 31:23

No, I think, I think we covered a lot of items. You know, I think, all of us are waiting to see what AI actually delivers, us included. And, you know, I think, the Future's bright and I can't wait to see what happens.

Tom Raftery: 31:38

Perfect. Super. Andy, if people would like to know more about yourself or any of the things we discussed on the podcast today, where would you have me direct them?

Andy Kohm: 31:47

Yeah, so they can go hit my LinkedIn profile. Happy to connect with them. Our website is www.my SCIP.com. or www SCIP.AI, both of those will take 'em to our website and, um, they can learn more about SCIP and we have some blog posts and we continue to, you know, write articles about, the supply chain and where it's going.

Tom Raftery: 32:09

Fantastic. Great, Andy, that's been really interesting. Thanks a million for coming on the podcast today.

Andy Kohm: 32:14

Yes. Thanks for having me.

Tom Raftery: 32:16

Okay. Thanks everyone for listening to this episode of the Resilient Supply Chain Podcast with me, Tom Raftery. Every week, thousands of senior supply chain and sustainability leaders tune in to learn what's next in resilience, innovation, and transformation. If your organisation wants to reach this influential global audience, the people shaping the future of supply chains, consider partnering with the show. Sponsorship isn't just brand visibility, it's thought leadership, credibility, and direct engagement with the decision makers driving change. To explore how we can spotlight your story or your solutions, connect with me on LinkedIn or drop me an email at Tom at tom Raftery dot com. Let's collaborate to build smarter, more resilient, more sustainable supply chains together. Thanks for tuning in, and I'll catch you all in the next episode.